Related Posts
How important is QBI for REG?
Is JS7 & L4 same ?? Any idea ??
Additional Posts in Jobs in STEM
New to Fishbowl?
Download the Fishbowl app to
unlock all discussions on Fishbowl.
unlock all discussions on Fishbowl.
How important is QBI for REG?
Is JS7 & L4 same ?? Any idea ??
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Download the Fishbowl app to unlock all discussions on Fishbowl.
Copy and paste embed code on your site

Scan your QR code to download
Fishbowl app on your mobile

I don't. Although I routinely spot errors from AI when I'm forced to review such. I feel the use of AI is a crutch providing laziness in critical thinking and reliance upon knowledge being summarized being plagiarized (Is not the answer normally found in the details?). Both of which, as any Scientist should agree, are abhorrent practices that do nothing to advance our species or science in general. AI is not the answer for us to evolve. Nor should it be allowed to affect our advancement. Look at all the strife it is causing already, and that only during its infancy.
I’ve noticed they’re only convincing when you’re working on things outside your realm of expertise. The trick is to look at the source material or prompt for it to provide references and source material so that you can look it up and confirm it is understanding things properly.
Not very often now that I've gotten better at prompting. The occasional hallucination will sneak in every once in a while, but I always catch it before the output ever leaves the chat window. A couple stern words (i.e. all caps expletives) is usually all it takes to fix whatever went wrong.
I truthfully have not used AI much in my work for this reason. I find that it does make mistakes often. It isn't perfect though and probably never will be.
Chief
AI is getting better each week. It’s going to be a pain if you try to verify everything. There’s a point where you are going to have to fully trust it.
Chief
I treat AI output like a draft assistant rather than a final answer. I still review everything carefully because convincing errors are exactly what make AI risky when people stop checking.
Not sure about the other model, since 4.6 Claude, I have never found anything off, even a bit. I do validate the output before using it but I think validation has become a waste of time.